File size: 17,752 Bytes
f50d964
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
from __future__ import annotations

import configparser
import pathlib
import typing
import os

import torch
import transformers
from torch.nn.utils.rnn import pad_sequence

from .config import LYRA_BAICHUAN_PARAM, LIB_SO_PATH
from .model import BaichuanModel
from .tokenization_baichuan import BaichuanTokenizer

class lyraBaichuan7B:
    def __init__(self, model_path, tokenizer_path=None, dtype='fp16', memopt_mode=0, arch='Ampere', cuda_version=12) -> None:
        self.model_path = model_path
        self.tokenizer_path = tokenizer_path
        self.dtype = dtype
        self.memopt_mode = memopt_mode
        self.arch = arch
        self.cuda_version = cuda_version

        self.model, self.tokenizer = self.load_model_and_tokenizer()
        print("Got model and tokenizer")

    def load_model_and_tokenizer(self):
        if self.tokenizer_path is None:
            tokenizer_path = self.model_path
        else:
            tokenizer_path = self.tokenizer_path

        print(f'Loading tokenizer from {tokenizer_path}')
        tokenizer = BaichuanTokenizer.from_pretrained(tokenizer_path)

        checkpoint_path = pathlib.Path(self.model_path)
        config_path = checkpoint_path / 'config.ini'

        if config_path.exists():
            # Read model params from config.
            cfg = configparser.ConfigParser()
            cfg.read(config_path)
            model_name = 'baichuan'
            inference_data_type = self.dtype
            if inference_data_type == None:
                inference_data_type = cfg.get(model_name, "weight_data_type")
            model_args = dict(
                head_num=cfg.getint(model_name, 'head_num'),
                size_per_head=cfg.getint(model_name, "size_per_head"),
                inter_size=cfg.getint(model_name, 'inter_size'),
                layer_num=cfg.getint(model_name, "num_layer"),
                rotary_embedding_dim=cfg.getint(model_name, 'rotary_embedding'),
                layernorm_eps=cfg.getfloat(model_name, 'layernorm_eps'),
                vocab_size=cfg.getint(model_name, "vocab_size"),
                start_id=cfg.getint(model_name, "start_id"),
                end_id=cfg.getint(model_name, "end_id"),
                weights_data_type=cfg.get(model_name, "weight_data_type"),
                tensor_para_size=cfg.getint(model_name, "tensor_para_size"),
                inference_data_type=inference_data_type)
        else:
            inference_data_type = self.dtype
            if inference_data_type == None:
                inference_data_type = LYRA_BAICHUAN_PARAM.weights_data_type
            model_args = dict(head_num=LYRA_BAICHUAN_PARAM.num_heads,
                              size_per_head=LYRA_BAICHUAN_PARAM.size_per_head,
                              inter_size=LYRA_BAICHUAN_PARAM.inter_size,
                              layer_num=LYRA_BAICHUAN_PARAM.num_layers,
                              rotary_embedding_dim=LYRA_BAICHUAN_PARAM.rotary_embedding,
                              layernorm_eps=LYRA_BAICHUAN_PARAM.layernorm_eps,
                              vocab_size=LYRA_BAICHUAN_PARAM.vocab_size,
                              start_id=LYRA_BAICHUAN_PARAM.start_id or tokenizer.bos_token_id,
                              end_id=LYRA_BAICHUAN_PARAM.end_id or tokenizer.eos_token_id,
                              weights_data_type=LYRA_BAICHUAN_PARAM.weights_data_type,
                              tensor_para_size=LYRA_BAICHUAN_PARAM.tensor_para_size,
                              inference_data_type=inference_data_type)

        # update common parameters
        
        # Load the C++ model into Pytorch model.
        sm = "sm80"
        
        if self.arch == "Ampere":
            sm = "sm80"
        elif self.arch == "Volta":
            sm = "sm70"
        else:
            raise Exception(f"unsupported arch: {self.arch}")

        cu = 'cu11'
        if self.cuda_version == 11:
            cu = 'cu11'
        elif self.cuda_version == 12:
            cu = 'cu12'
        else:
            raise Exception(f"unsupported cuda version: {self.cuda_version}")
        
        lib_path = pathlib.Path(__file__).parent / "ftlib" / f"libth_transformer_{sm}_{cu}.so"
        
        model_args.update(dict(
            lib_path=lib_path,
            model_path=os.path.join(self.model_path, "1-gpu-fp16.bin"),
            max_seq_len=0,  # for position seq embedding
            pipeline_para_size=LYRA_BAICHUAN_PARAM.pipeline_para_size,
            use_gptj_residual=LYRA_BAICHUAN_PARAM.use_gptj_residual,
            memopt_mode=self.memopt_mode
        ))

        print('[FT][INFO] Load Our FT Highly Optimized Baichuan-7B model')
        for k, v in model_args.items():
            print(f' - {k.ljust(25, ".")}: {v}')

        # Check sanity and consistency between the model and tokenizer.
        checklist = ['head_num', 'size_per_head', 'vocab_size', 'layer_num',
                     'tensor_para_size', 'tensor_para_size', 'weights_data_type']
        if None in [model_args[k] for k in checklist]:
            none_params = [p for p in checklist if model_args[p] is None]
            print(f'[FT][WARNING] Found None parameters {none_params}. They must '
                  f'be provided either by config file or CLI arguments.')
        if model_args['start_id'] != tokenizer.bos_token_id:
            print('[FT][WARNING] Given start_id is not matched with the bos token '
                  'id of the pretrained tokenizer.')
        if model_args['end_id'] not in (tokenizer.pad_token_id, tokenizer.eos_token_id):
            print('[FT][WARNING] Given end_id is not matched with neither pad '
                  'token id nor eos token id of the pretrained tokenizer.')

        print(f'Loading model from {self.model_path}')
        model = BaichuanModel(**model_args)
        return model, tokenizer

    def generate(self, prompts: typing.List[str] | str,
                 output_length: int = 512,
                 beam_width: int = 1,
                 top_k: typing.Optional[torch.IntTensor] = 1,
                 top_p: typing.Optional[torch.FloatTensor] = 1.0,
                 beam_search_diversity_rate: typing.Optional[torch.FloatTensor] = 0.0,
                 temperature: typing.Optional[torch.FloatTensor] = 1.0,
                 len_penalty: typing.Optional[torch.FloatTensor] = 0.0,
                 repetition_penalty: typing.Optional[torch.FloatTensor] = 1.0,
                 presence_penalty: typing.Optional[torch.FloatTensor] = None,
                 min_length: typing.Optional[torch.IntTensor] = None,
                 bad_words_list: typing.Optional[torch.IntTensor] = None,
                 do_sample: bool = False,
                 return_output_length: bool = False,
                 return_cum_log_probs: int = 0):
        #
        if isinstance(prompts, str):
            prompts = [prompts, ]

        inputs = prompts

        batch_size = len(inputs)
        ones_int = torch.ones(size=[batch_size], dtype=torch.int32)
        ones_float = torch.ones(size=[batch_size], dtype=torch.float32)

        # we must encode the raw prompt text one by one in order to compute the length of the original text.
        input_token_ids = [self.tokenizer(text, return_tensors="pt").input_ids.int().squeeze() for text in inputs]
        input_lengths = torch.IntTensor([len(ids) for ids in input_token_ids])
        # after got the length of each input text tokens. we can batchfy the input list to a tensor. padding the right.
        input_token_ids = pad_sequence(input_token_ids, batch_first=True, padding_value=self.tokenizer.eos_token_id)

        random_seed = None
        if do_sample:
            random_seed = torch.randint(0, 262144, (batch_size,), dtype=torch.long)

        outputs = self.model(start_ids=input_token_ids,
                             start_lengths=input_lengths,
                             output_len=output_length,
                             beam_width=beam_width,
                             top_k=top_k * ones_int,
                             top_p=top_p * ones_float,
                             beam_search_diversity_rate=beam_search_diversity_rate * ones_float,
                             temperature=temperature * ones_float,
                             len_penalty=len_penalty * ones_float,
                             repetition_penalty=repetition_penalty * ones_float,
                             random_seed=random_seed,
                             return_output_length=return_output_length,
                             return_cum_log_probs=return_cum_log_probs)

        if return_cum_log_probs > 0:
            outputs = outputs[0]  # output_token_ids.

        # Slice the generated token ids of the 1st beam result.
        # output = input tokens + generated tokens.
        output_token_ids = [out[0, length:].cpu()
                            for out, length in zip(outputs, input_lengths)]

        output_texts = self.tokenizer.batch_decode(
            output_token_ids, skip_special_tokens=True)

        return output_texts
    
class lyraBaichuan13B:
    def __init__(self, model_path, tokenizer_path=None, dtype='fp16', memopt_mode=0, arch='Ampere', cuda_version=12) -> None:
        self.model_path = model_path
        self.tokenizer_path = tokenizer_path
        self.dtype = dtype
        self.memopt_mode = memopt_mode
        self.arch = arch
        self.cuda_version = cuda_version

        self.model, self.tokenizer = self.load_model_and_tokenizer()
        print("Got model and tokenizer")

    def load_model_and_tokenizer(self):
        if self.tokenizer_path is None:
            tokenizer_path = self.model_path
        else:
            tokenizer_path = self.tokenizer_path

        print(f'Loading tokenizer from {tokenizer_path}')
        tokenizer = BaichuanTokenizer.from_pretrained(tokenizer_path)

        checkpoint_path = pathlib.Path(self.model_path)
        config_path = checkpoint_path / 'config.ini'

        if config_path.exists():
            # Read model params from config.
            cfg = configparser.ConfigParser()
            cfg.read(config_path)
            model_name = 'baichuan'
            inference_data_type = self.dtype
            if inference_data_type == None:
                inference_data_type = cfg.get(model_name, "weight_data_type")
            model_args = dict(
                head_num=cfg.getint(model_name, 'head_num'),
                size_per_head=cfg.getint(model_name, "size_per_head"),
                inter_size=cfg.getint(model_name, 'inter_size'),
                layer_num=cfg.getint(model_name, "num_layer"),
                rotary_embedding_dim=0,
                layernorm_eps=cfg.getfloat(model_name, 'layernorm_eps'),
                vocab_size=cfg.getint(model_name, "vocab_size"),
                start_id=cfg.getint(model_name, "start_id"),
                end_id=cfg.getint(model_name, "end_id"),
                weights_data_type=cfg.get(model_name, "weight_data_type"),
                tensor_para_size=cfg.getint(model_name, "tensor_para_size"),
                inference_data_type=inference_data_type)
        else:
            inference_data_type = self.dtype
            if inference_data_type == None:
                inference_data_type = LYRA_BAICHUAN_PARAM.weights_data_type
            model_args = dict(head_num=LYRA_BAICHUAN_PARAM.num_heads,
                              size_per_head=LYRA_BAICHUAN_PARAM.size_per_head,
                              inter_size=LYRA_BAICHUAN_PARAM.inter_size,
                              layer_num=LYRA_BAICHUAN_PARAM.num_layers,
                              rotary_embedding_dim=0,
                              layernorm_eps=LYRA_BAICHUAN_PARAM.layernorm_eps,
                              vocab_size=LYRA_BAICHUAN_PARAM.vocab_size,
                              start_id=LYRA_BAICHUAN_PARAM.start_id or tokenizer.bos_token_id,
                              end_id=LYRA_BAICHUAN_PARAM.end_id or tokenizer.eos_token_id,
                              weights_data_type=LYRA_BAICHUAN_PARAM.weights_data_type,
                              tensor_para_size=LYRA_BAICHUAN_PARAM.tensor_para_size,
                              inference_data_type=inference_data_type)

        # update common parameters
        # Load the C++ model into Pytorch model.
        sm = "sm80"
        
        if self.arch == "Ampere":
            sm = "sm80"
        elif self.arch == "Volta":
            sm = "sm70"
        else:
            raise Exception(f"unsupported arch: {self.arch}")

        cu = 'cu11'
        if self.cuda_version == 11:
            cu = 'cu11'
        elif self.cuda_version == 12:
            cu = 'cu12'
        else:
            raise Exception(f"unsupported cuda version: {self.cuda_version}")
        
        lib_path = pathlib.Path(__file__).parent / "ftlib" / f"libth_transformer_{sm}_{cu}.so"
        model_args.update(dict(
            lib_path=lib_path,
            model_path=os.path.join(self.model_path, "1-gpu-fp16.bin"),
            max_seq_len=0,  # for position seq embedding
            pipeline_para_size=LYRA_BAICHUAN_PARAM.pipeline_para_size,
            use_gptj_residual=LYRA_BAICHUAN_PARAM.use_gptj_residual,
            memopt_mode=self.memopt_mode
        ))

        print('[FT][INFO] Load Our FT Highly Optimized Baichuan-13B model')
        for k, v in model_args.items():
            print(f' - {k.ljust(25, ".")}: {v}')

        # Check sanity and consistency between the model and tokenizer.
        checklist = ['head_num', 'size_per_head', 'vocab_size', 'layer_num',
                     'tensor_para_size', 'tensor_para_size', 'weights_data_type']
        if None in [model_args[k] for k in checklist]:
            none_params = [p for p in checklist if model_args[p] is None]
            print(f'[FT][WARNING] Found None parameters {none_params}. They must '
                  f'be provided either by config file or CLI arguments.')
        if model_args['start_id'] != tokenizer.bos_token_id:
            print('[FT][WARNING] Given start_id is not matched with the bos token '
                  'id of the pretrained tokenizer.')
        if model_args['end_id'] not in (tokenizer.pad_token_id, tokenizer.eos_token_id):
            print('[FT][WARNING] Given end_id is not matched with neither pad '
                  'token id nor eos token id of the pretrained tokenizer.')

        print(f'Loading model from {self.model_path}')
        model = BaichuanModel(**model_args)
        return model, tokenizer

    def generate(self, prompts: typing.List[str] | str,
                 output_length: int = 512,
                 beam_width: int = 1,
                 top_k: typing.Optional[torch.IntTensor] = 1,
                 top_p: typing.Optional[torch.FloatTensor] = 1.0,
                 beam_search_diversity_rate: typing.Optional[torch.FloatTensor] = 0.0,
                 temperature: typing.Optional[torch.FloatTensor] = 1.0,
                 len_penalty: typing.Optional[torch.FloatTensor] = 0.0,
                 repetition_penalty: typing.Optional[torch.FloatTensor] = 1.0,
                 presence_penalty: typing.Optional[torch.FloatTensor] = None,
                 min_length: typing.Optional[torch.IntTensor] = None,
                 bad_words_list: typing.Optional[torch.IntTensor] = None,
                 do_sample: bool = False,
                 return_output_length: bool = False,
                 return_cum_log_probs: int = 0):
        #
        if isinstance(prompts, str):
            prompts = [prompts, ]

        inputs = prompts

        batch_size = len(inputs)
        ones_int = torch.ones(size=[batch_size], dtype=torch.int32)
        ones_float = torch.ones(size=[batch_size], dtype=torch.float32)

        # we must encode the raw prompt text one by one in order to compute the length of the original text.
        input_token_ids = [self.tokenizer(text, return_tensors="pt").input_ids.int().squeeze() for text in inputs]
        input_lengths = torch.IntTensor([len(ids) for ids in input_token_ids])
        # after got the length of each input text tokens. we can batchfy the input list to a tensor. padding the right.
        input_token_ids = pad_sequence(input_token_ids, batch_first=True, padding_value=self.tokenizer.eos_token_id)

        random_seed = None
        if do_sample:
            random_seed = torch.randint(0, 262144, (batch_size,), dtype=torch.long)

        outputs = self.model(start_ids=input_token_ids,
                             start_lengths=input_lengths,
                             output_len=output_length,
                             beam_width=beam_width,
                             top_k=top_k * ones_int,
                             top_p=top_p * ones_float,
                             beam_search_diversity_rate=beam_search_diversity_rate * ones_float,
                             temperature=temperature * ones_float,
                             len_penalty=len_penalty * ones_float,
                             repetition_penalty=repetition_penalty * ones_float,
                             random_seed=random_seed,
                             return_output_length=return_output_length,
                             return_cum_log_probs=return_cum_log_probs)

        if return_cum_log_probs > 0:
            outputs = outputs[0]  # output_token_ids.

        # Slice the generated token ids of the 1st beam result.
        # output = input tokens + generated tokens.
        output_token_ids = [out[0, length:].cpu()
                            for out, length in zip(outputs, input_lengths)]

        output_texts = self.tokenizer.batch_decode(
            output_token_ids, skip_special_tokens=True)

        return output_texts